• Horizon AI
  • Posts
  • How to Keep Your Conversations with Claude and Copilot as Private as Possible ๐Ÿ›ก๏ธ

How to Keep Your Conversations with Claude and Copilot as Private as Possible ๐Ÿ›ก๏ธ

GPT-6 Astra Is Finally Here ๐Ÿ”ฅ

In partnership with

Welcome to another edition of Horizon AI,

A few weeks ago, we looked at the privacy settings you can use to keep your ChatGPT and Gemini conversations more private. Today, we're continuing that guide with Claude and Microsoft Copilot, covering the settings you can change to limit data collection, manage shared conversations, and control what each AI remembers about you.

Letโ€™s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

  • OpenAI Releases GPT-6 Astra as It Declares the Start of the AGI Era

  • Google Introduces WeatherNext 3

  • AI Tutorial: How to Keep Your Claude and Copilot Conversations More Private

  • AI Tools to check out

  • AI Findings/Resources

  • The Latest in AI and Tech ๐Ÿ’ก

AI News

OPENAI

OpenAI Releases GPT-6 Astra as It Declares the Start of the "AGI Era"

OpenAI has released GPT-6 Astra, "the world's most intelligent and aligned model," with cofounder and President Greg Brockman calling it a "generational leap" and suggesting it could eventually be seen as the arrival of AGI.

Details:

  • The company said Astra was built on its largest-ever training run, using more than 100,000 GPUs at its Stargate site in Texas.

  • The model performs at the highest level on benchmarks including FrontierMath Tier 1 (97.6%), which tests mathematical ability; ARC-AGI 3 (99.9%) for reasoning ability; ExploitBench (100%) for cybersecurity; and DeepSWE v1.1 (74.1%) for software engineering.

  • It also shows significant improvements in areas like using a computer, and professional work like creating presentations and spreadsheets.

  • When it comes to pricing, GPT-6 Astra will be 2.5 times more expensive than GPT-5.6 Sol, at $10 per million input tokens and $50 per million output tokens (the same pricing as Anthropic's Fable 5.1 model).

  • Brockman also hinted that OpenAI, and the industry as a whole, could soon move away from per-token pricing for AI models: "the price per task is what matters," he said.

The new flagship model will initially be available to a limited number of organizations and, over the coming days, will become available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenAI API, Microsoft Azure, and AWS Bedrock.

TOGETHER WITH WANDERBOAT

Say What You Actually Want

"Dog-friendly, cozy, golden-hour patio, serves natural wine." Try typing that into a normal map app and watch it break. Most search tools force you into rigid keywords and ignore what actually makes a place right for you. Wanderboat works differently. You describe exactly what you want, as specific or chaotic as it sounds in your head, and it finds the match.

  • Type your request in your own words, no keyword guessing

  • Combine as many details as you want, at once

  • Book tables, order food, reserve hotels, or buy tickets directly in Wanderboat

  • Get results built around your exact vibe, not a generic list

The pickiest request you've got? Wanderboat can handle it and help you make the plan happen.

GOOGLE

Google Introduces WeatherNext 3

Google has released WeatherNext 3, its latest AI model for global weather forecasting, with major improvements in forecast accuracy, resolution, and frequency.

Details:

  • WeatherNext 3 is the most advanced and accurate global weather model to date, outperforming leading AI and traditional forecasting systems, according to independent live evaluations by Brightband.

  • It can produce forecasts at a resolution of up to 5 km, compared with the wider areas covered by many existing AI models.

  • Rain forecasting improves by 60% compared with WeatherNext 2, while forecasts can now be generated hourly instead of every six hours.

  • The model is able to forecast more frequently because it can ingest weather satellite data collected in real time on an hourly basis.

WeatherNext 3 will begin powering weather information across Google products and will also be available to users and researchers through Googleโ€™s cloud platforms.

AI Tutorial

How to Keep Your Claude and Copilot Conversations More Private

  1. Disable AI training: On the Copilot website or desktop apps, click your account name at the bottom of the left pane, select Settings, and then Privacy. Turn off the switches for both "Training on conversation activity" and "Training on voice conversations."

  2. Delete links to shared conversations: Go to Settings > Privacy > Manage Shared Links and revoke any shared conversation links you no longer want accessible.

  3. Turn off memory: Go to Settings > Memory and turn off "Microsoft usage data" and "Personalization and memory" if you don't want Copilot using information from your Microsoft services or remembering details about you.

  4. Remove connected services: Go to Settings > Connectors and turn off access to services such as OneDrive, Outlook, Google Drive, Gmail, or Google Calendar if you don't want Copilot accessing data from them.

  1. Use Incognito chats: Click the ghost icon in the upper-right corner of a new chat to start an Incognito conversation when you don't want Claude to store it.

  2. Control how you share a conversation: When you click the Share icon, you can select the "Only people invited" setting, which keeps your Claude conversation completely private so that only you and explicitly invited members can access it.

  1. Delete shared links: Go to Settings > Privacy > Shared chats > Manage and delete any public conversation links you no longer want accessible.

  2. Disable AI training: Go to Settings > Privacy and make sure "Help improve our AI models" is turned off.

  3. Turn off memory: Go to Settings > Memory and turn off "Generate memory from chats." You can then pause the memory feature to disable it but retain any existing memories, or reset the memory to turn it off and remove any existing memories.

AI Tools to check out

โœ… Perplexity: a free AI-powered answer engine that provides accurate, trusted, and real-time answers to any question.

๐Ÿ–ฑ๏ธ RenameClick: a local-first desktop app for macOS and Windows that renames, organizes, and searches files by their actual content

๐Ÿงฉ Puzzle: AI accounting software for startups & accounting firms.

๐Ÿš€ Soloop: The one-person company OS that gives solo founders an AI founding team to grow the product they built.

๐Ÿฆพ OpenWorker: an AI agent that runs on your own computer, works in your everyday tools, and turns what you ask into finished deliverables

AI Findings/Resources

๐Ÿค” 91% of professionals say their firm still falls short on AI โ€“ how to fix that

The latest in AI and Tech

The company is offering major discounts on its Muse Spark model to users who share their prompts and outputs for future model training. Contributor pricing: $0.10 per million input tokens (vs. $1.25 standard) and $0.20 per million output tokens (vs. $4.25 standard)

Mira Murati's Thinking Machines Lab is reportedly in talks to raise $1 billion at a valuation of at least $40 billion, with existing backer Accel potentially leading the round. The AI startup previously raised $2 billion at a $12 billion valuation.

Hugging Face will remain an open platform while gaining access to Nvidiaโ€™s resources and computing infrastructure.

ChatGPT, Claude, Grok, and Gemini experienced overlapping service disruptions on Thursday morning. The cause of the simultaneous outage remains unclear.

Thatโ€™s a wrap!

Thanks for sticking with us to the end! Letโ€™s stay connected on LinkedIn and Twitter.

We'd love to hear your thoughts on today's email!

Your feedback helps us improve our content

Login or Subscribe to participate in polls.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina ๐Ÿ‘ฉ๐Ÿปโ€๐Ÿ’ป